Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Meta feed video ads start at 1 second; so does Sume Timeline
Both Meta feed video pages list a 1-second minimum. Sume Timeline accepts 1 second of output and bills one rounded-up minute, $0.10, though models start at 4 s.
GPT Image 2.5 high: 1920x1080 costs less than 1024x1024
On fal, GPT Image 2.5 Flare high is $0.0396 at 1920x1080 and $0.05268 at 1024x1024. Why the bigger frame is cheaper and how to plan around it.
Go: a context timeout stops waiting on a Sume job, not the job
context.WithTimeout cancels your status read, never the generation. A 29-line Go loop shows the DeadlineExceeded branch, and why the job id must be stored.
Claude Code MCP scope: local, project or user for the Sume server
Use project scope for a shared .mcp.json with the Sume URL only, user scope for your own machine, and keep API keys out of shared files.
Sume run webhook: two request_ids, which one to dedupe on
Dedupe a Sume run webhook on the envelope request_id, which equals run_id and is stable across retries. Ignore payload.request_id, a correlation id.
Python: check Omni Flash 1.1 limits against /v1/videos/models first
Google's Omni runs a sync call and extends in 10-second steps up to 40 seconds. Sume's gemini-omni-flash-1.1 takes 3 to 10 seconds per job. A preflight check.
Alibaba task_status UNKNOWN: map DashScope states to Sume job states
Model Studio reports PENDING, RUNNING, SUCCEEDED, FAILED, CANCELED and UNKNOWN. Sume has five job states and no twin for UNKNOWN. A small map and a rule.
Pinterest Showcase video: 3 to 60 seconds, 2:3, 1000x1500 on Sume
Pinterest's spec page lists 3-60 seconds, 2:3 and 1000 x 1500 for Showcase video. Sume Timeline can render 1000x1500 and a 3-second cut costs the same $0.10.
Omni 40-second ad: resend the packshot at every 10-second step
Google extends Omni clips to 40 s in 10 s steps with 10 s of prior context. On Sume, four 720p clips with the packshot as a reference cost $5.00.
After imagen-4.0-generate-001: read n from endpoints, use webhook mode
Do not assume your old image count carries over. Read supported_parameters.n from the Sume endpoints route, then submit in webhook mode and wait for.
Pinterest Collections video: standard 4 s to 15 min spec, 2:3 hero
Pinterest says Collections follow its standard video spec: 4 seconds to 15 minutes, 2 GB, with a 2:3 ratio. Here is what a Sume render has to match.
Can Gemini Omni Flash speak my script? It takes no audio input on Sume
On Sume, Gemini Omni Flash 1.1 takes image and video references but no audio reference. It makes its own sound. For your script in a face, use avatar video.
LTX-2.5 Fast vs Wan 3.0: price per second at 720p and 1080p
LTX-2.5 Fast on fal is $0.09 at 720p and $0.13 at 1080p; Wan 3.0 lists $0.10 and $0.20 at Alibaba, $0.125 and $0.25 on Sume. A 10-second compare.
Go: poll a Sume job and stop on any terminal state, even a new one
A Go status loop that names completed, failed and canceled, stops on terminal=true for anything else, and never sleeps less than next_poll_after_seconds.
HappyHorse 1.1 takes a seed on Cloudflare; Sume video models do not
Cloudflare's HappyHorse 1.1 accepts a seed up to 2,147,483,647 and an optional watermark. No Sume v1 video model accepts a seed; here is the workaround.
YouTube 480p at 2.5 Mbps and 360p at 1 Mbps: size of a 60-second draft
YouTube lists 2.5 Mbps for 480p and 1 Mbps for 360p SDR uploads. A 60-second clip at those rates is about 18.75 MB and 7.5 MB; Sume sets no bitrate itself.
Qwen Image 2.1 Pro API: the model card lists no Pro variant
Searching for a Qwen Image 2.1 Pro API? The 2.1 model card names no Pro variant and no hosted endpoint. Here is what Sume lists for Qwen today.
sume/auto is missing from GET /v1/images/models: hardcode it
Sume Image Router Auto never appears in the model list. Discovery code that only reads GET /v1/images/models will never offer sume/auto. How to handle it.
TTS output formats: Gemini 24 kHz WAV vs Sume 44.1 kHz MP3 default
Gemini 3.8 TTS returns 24 kHz mono 16-bit WAV, or headerless L16 when streaming. Sume TTS defaults to 44.1 kHz 128 kbps MP3 and offers WAV and raw options.
MAI custom voice SSML (ttsembedding) vs Sume voice ids: what is gated
Microsoft's MAI-Voice-2.1 custom voice sits behind Limited Access Review and uses a speakerProfileId in SSML. Sume picks voices by id, with no clone widget.
OpenAI Image API vs Sume Image API for GPT Image 2.5: what differs
Same model family, different request surface. Moderation, revised_prompt, streaming, output_compression and masks compared between OpenAI direct and Sume.
Crop 9:16 to 16:9 for TikTok horizontal: height 0.3164 is 1080x608
A 1080x1920 clip cropped to 16:9 keeps 1080x608, which clears TikTok's 960x540 horizontal minimum but uses a third of the frame. Numbers and request inside.
Ruby Net::HTTP: POST /v1/videos with callback_url, for a Sora port
A 15-line Net::HTTP submit for /v1/videos that sends an HTTPS callback_url and an Idempotency-Key, checks for 202, and stores the id before any callback.
A lost response on a 30 s Kling motion control submit: reuse the key
If a network timeout hides the reply to a 30-second Kling motion control submit, a new Idempotency-Key reserves $4.725 again. Reuse the key. Python retry loop.
After Sora: keep the recipe, not the prompt, as your stable layer
OpenAI named no replacement for the Sora Videos API. Put house style and output rules in a Sume Format so the next model change does not rewrite your code.
Anime-style ad with Seedance 2.5: use a character sheet as references
Send your mascot's character sheet as input_references to seedance-2.5 on Sume and test a 6-second anime-style ad at 480p before you pay for 720p.
Anime-style video prompt for Gemini Omni: cel look and limits
Get a hand-drawn anime look from Gemini Omni with art-direction words, a reference image option and a 5-second Sume request. Prices for drafts and finals.
Avatar video quality tiers: test one script at all three
How to choose between standard, plus and max in Sume Avatar 1.0: render one script three times, compare on a checklist, and keep the cheapest tier that passes.
Boutique dress try-on video from one garment photo on Sume
A small boutique can preview one dress on a person in a vertical video with the sume-virtual-try-on Format. What to send, what it is not, and the spend cap.
Budget for 500 AI video clips a month: plan fee and slots
500 ten-second 768p clips are $500 of usage. That is $500 on Pro, $480 on Startup and $400 on Scale, with 125, 63 and 25 waves of concurrency.
Car detailing before-after clip: first and last frame on Seedance 2
Use a dirty-car photo as the first frame and a finished-car photo as the last frame in one Seedance 2 job on Sume. Limits, request body, and what to check.
Cartesia credits to dollars vs Sume TTS: per minute and per hour
Cartesia plans work out to $0.0374 to $0.05 per 1,000 characters; Sume TTS is $0.0475 (Cartesia list x 1.25). 1,000 minutes: $29.90 to $40 vs $38.
Same character in three AI video shots: Omni reference image
Keep one character across separate Gemini Omni clips by sending the same reference image with every request on Sume. Request shape, prompt tags, 3-shot cost.
ChatGPT Images 2.5 Poster and Merch templates as API prompts
ChatGPT Images 2.5 added Poster and Merch templates. The API has no template field, so write the layout into the prompt and call openai/gpt-image-2.5 on Sume.
How long do 100 AI video clips take? Concurrency slots by plan
100 jobs take ceil(100 / slots) waves: 100 on Free, 25 on Pro, 13 on Startup, 5 on Scale. Multiply by one job's time. Math and a runnable snippet.
Content ID on a Short over one minute blocks it globally: keep a cut
YouTube says a video over one minute with an active Content ID claim is blocked globally. Keep a 59-second video-trim cut of every long Short as a fallback.
ElevenLabs plans vs Sume TTS: cost per million characters
ElevenLabs plans work out to about $0.022 per 1,000 v4 characters, so 1 million characters is about $22. Sume TTS is $47.50 per million.
Film noir black-and-white AI video prompt for Gemini Omni
A black-and-white noir prompt for Gemini Omni Flash: lighting words, a no-dialogue sound line and a 10-second Sume request with prices for 720p and 1080p.
Foggy forest morning clip with ambient sound: Gemini Omni, 10s
Prompt a 10-second foggy forest mood clip in Gemini Omni with ambient birds and drips. Sume request, why people-free scenes are easier, and cost per resolution.
Build a run status chip from the Format events phase timeline
Sume has no SSE stream for Format runs. Poll status_url and events_url to show queued, preparing, running and finalizing in your UI without faking progress.
Forward a finished Sume run to team chat with a signed webhook
A standard-library Python receiver that verifies the sume-v1 signature, rejects an empty secret, and forwards primary_output_url when a Format run ends.
frame_images and input_references together: frame_images wins
Send both and Sume runs image-to-video, not reference-to-video. How to pick one field per request and check the model's catalog entry first.
Gemini 3.8 Flash TTS context caching vs Sume transcripts
Gemini 3.8 Flash TTS prices cached input at $0.125 per million tokens until 2026-12-31. Sume sends the full transcript on each job. When caching matters.
Gemini TTS short pause and breath tags vs Sume sentence segments
Gemini 3.8 Flash TTS takes inline tags like short pause and breath. Sume lists no inline tags; here is how to control pacing with segments and speed.
Gemini TTS streams raw PCM; Sume returns a finished wav or raw file
Gemini 3.8 Flash TTS streams headerless audio/l16 at 24 kHz. Sume returns a hosted file from an async job. Wrap the PCM in WAV and match the formats.
Gemini Omni Flash 1.1 API cost per clip, 360p to 4K
Gemini Omni Flash 1.1 on Sume is $0.0375 to $0.375 per second. An 8-second 720p clip is $1.00; 100 clips are $100. Full table by resolution.
MiniMax H3 Max API price: Sume 1.25x vs the fal list
H3 Max costs $0.0625 to $0.20 per second on Sume (list x 1.25). Per-clip and per-100-clip math against fal's listed $0.05 per second.
HeyGen bills 2K at 3x its 768p rate: price a resolution bump on Sume
HeyGen bills 2K at 3x its 768p rate. On Sume, read pricing_skus and supported_resolutions from /v1/videos/models and show an estimate before a 2K toggle.